A method for scheduling unmanned aerial vehicle flight missions

Abstract

The present invention discloses a method for scheduling unmanned aerial vehicle flight tasks, comprising the following steps: (1) establishing decision variables; (2) setting an objective function; (3) establishing constraint conditions; (4) encoding and generating initial chromosomes; (5) setting parameters, and generating an initial population by optimizing with a variable neighborhood heuristic; (6) calculating a selection probability according to individual fitness, and selecting a parent generation according to the probability; (7) performing a mutation crossover operation on two randomly selected parent generations according to the size of an adaptive crossover probability; (8) selecting the original parent generation and the newly generated next generation as individuals of the next generation according to the fitness ranking; (9) adding the offspring obtained in step (8) to the population, and repeating steps (6)-(8) until the number of iterations reaches a maximum. The present invention improves the efficiency of scheduling unmanned aerial vehicle flight tasks.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) mission scheduling, and in particular to a UAV flight mission scheduling method. Background Art

[0002] As one of the rapidly developing technologies in recent years, drone technology has been widely used in various fields such as logistics and distribution. One of the key problems to be solved is the flight mission scheduling of multiple drones.

[0003] UAV flight mission scheduling is to reduce the time and delivery cost of tasks by reasonably allocating tasks through mutual cooperation. It is a combinatorial optimization problem with multiple objectives and multiple constraints. The current multi-UAV flight mission scheduling methods mainly include distributed group collaborative dynamic task allocation method based on extended contract network protocol, mixed integer linear programming task allocation method, semi-random Q learning algorithm, adaptive genetic method, etc. The above methods have their own advantages and disadvantages when applied to path planning. Genetic algorithms have strong global search capabilities and weak local search capabilities. In the genetic algorithm solution process, the initial selection is relatively random, the convergence speed is slow, excellent individuals are prone to change, and they are prone to fall into local optimality. Often, only suboptimal solutions can be obtained instead of optimal solutions. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a method for scheduling UAV flight missions, establish a cost function with the goal of minimizing the comprehensive flight cost based on management cost, transportation cost, and delay penalty; consider the UAV's range, load and other constraints, construct a multi-machine task allocation model, and incorporate a mixed initial population generation method of random and neighborhood optimization, a selection strategy based on linear sorting, an improved crossover mutation strategy, and an improved crossover probability calculation to solve the problems existing in the background technology.

[0005] Technical solution: A method for scheduling UAV flight missions according to the present invention comprises the following steps:

[0006] (1) Establish decision variables;

[0007] (2) Setting the objective function;

[0008] (3) Establish constraints;

[0009] (4) Encoding and generating initial chromosomes;

[0010] (5) Set parameters and use variable neighborhood heuristic optimization to generate the initial population;

[0011] (6) Calculate the selection probability based on individual fitness and select the parent generation based on the probability;

[0012] (7) Perform a mutation crossover operation on the two randomly selected parents according to the size of the adaptive crossover probability;

[0013] (8) The original parent generation and the newly generated next generation are selected as individuals of the next generation according to the fitness ranking;

[0014] (9) Add the offspring obtained in step (8) to the population and repeat steps (6)-(8) until the number of iterations reaches the maximum.

[0015] Furthermore, step (1) is as follows: Assume there are n drones, and let the set of drones be U, then , in, ; represents any one of the drones; there are m delivery demand points, each demand point corresponds to a delivery task, and the task set is T, then , in, j =1,2… m represents any one of the delivery tasks; the number of task demand points is greater than the number of drones, that is, m>n; the process of the drone returning to its original position after completing the flight mission is recorded as ; Then the decision variables of the task allocation model are The formula is as follows:

[0016] .

[0017] Furthermore, the formula of step (2) is as follows: The objective function F is set as:

[0018] ;

[0019] in, , are the weight coefficients of economic cost and delay penalty respectively, and the parameters satisfy ; For economic costs; for delayed punishment;

[0020] ;

[0021] ;

[0022] ;

[0023] in, Indicates drone The cost of transporting a unit distance at maximum load; Indicates drone Fly from the previous mission point to the mission point The length of the distance; Indicates drone go through The load-bearing capacity during the period; Indicates drone Maximum load; For drones Own weight; For drones Management cost per unit time; Indicates drone U i Flight speed;

[0024] ;

[0025] ;

[0026] ;

[0027] in, represents the penalty coefficient factor; t ij Indicates drone U i Execute the mission Arrive at the mission T j And the time to complete unloading; ET is the maximum allowed time for delivery.

[0028] Furthermore, in step (3), the constraints include:

[0029] Delivery task constraints:

[0030] ;

[0031] Load constraints:

[0032] ;

[0033] in, q i Indicates i The total load of the drone is less than the maximum load q max ; q j It means the task T j The weight of the cargo required.

[0034] Furthermore, in step (3), the constraints also include:

[0035] Distance constraint:

[0036] .

[0037] Furthermore, in step (5), the maximum number of evolution generations is set M , decision variables, objective function, constraints, and population size parameters.

[0038] Furthermore, step (6) is as follows: calculate the fitness of each chromosome according to the objective function, sort them from small to large in fitness, and calculate the selection probability according to the ranking order. P (i), the formula is as follows:

[0039] ;

[0040] in, i =1,2… N ; N is the number of individuals; j is the current algebra; M is the maximum evolutionary generation; all individuals get different selection probabilities according to their rankings; and then the parent generation is selected according to the individual's selection probability.

[0041] Furthermore, step (7) is as follows: Set the crossover probability p c , after selecting the crossover individual, a number between 0 and 1 is randomly generated. If the random number is greater than or equal to the crossover probability p c , then randomly select two positions G1 and G2 on the gene and invert the gene part between them. If it is less than the crossover probability p c , then find the gene position of G3 adjacent to G1 in another individual, and find two inversions at the original gene to form a new individual; the crossover probability p c The formula is as follows:

[0042] ;

[0043] ;

[0044] in, P c_min is the minimum crossover probability, P c_max is the maximum crossover probability, f is the individual fitness, f max is the maximum value of fitness, f min is the minimum value of fitness, f avg is the average fitness of all individuals in the current generation, A ,B is a constant.

[0045] An electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, any one of the methods for scheduling unmanned aerial vehicle flight missions is implemented.

[0046] A storage medium described in the present invention stores a computer program, and when the computer program is executed by a processor, it implements any one of the methods for scheduling unmanned aerial vehicle flight missions.

[0047] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: the method of the present invention is superior to the prior art in fitness function value, management cost, transportation cost, and delay penalty evaluation indicators. The present invention improves the efficiency of UAV flight mission scheduling, reduces time, solves the problem of being easily trapped in local areas, and the results are better. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is the coding principle diagram of the present invention;

[0049] Figure 2 It is a diagram of the variable neighborhood heuristic optimization principle of the present invention;

[0050] Figure 3 The present invention is a variation operation diagram;

[0051] Figure 4 It is a schematic diagram of the crossover strategy of the present invention. DETAILED DESCRIPTION

[0052] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0053] like Figure 1 As shown, an embodiment of the present invention provides a method for scheduling flight tasks of a UAV, comprising the following steps:

[0054] (1) Establish decision variables, as follows: The purpose of UAV flight mission scheduling is to determine the mission objectives of each UAV and plan the reasonable execution order of the missions, and obtain the corresponding order relationship between each UAV and the mission target point. Assuming there are n UAVs, and the set of UAVs is U, then , in, ; represents any one of the drones; there are m delivery demand points, each demand point corresponds to a delivery task, and the task set is T, then , in, j =1,2… mrepresents any one of the delivery tasks; the number of task demand points is greater than the number of drones, that is, m>n; the process of the drone returning to its original position after completing the flight mission is recorded as ; Then the decision variables of the task allocation model are The formula is as follows:

[0055] .

[0056] (2) Set the objective function as follows: The objective function F is set as:

[0057] ;

[0058] in, , are the weight coefficients of economic cost and delay penalty respectively, and the parameters satisfy ; For economic costs; for delayed punishment;

[0059] ;

[0060] ;

[0061] ;

[0062] in, Indicates drone The cost of transporting a unit distance at maximum load; Indicates drone Fly from the previous mission point to the mission point The length of the distance; Indicates drone go through The load-bearing capacity during the period; Indicates drone Maximum load; For drones Own weight; For drones Management cost per unit time; Indicates drone U i Flight speed;

[0063] ;

[0064] ;

[0065] ;

[0066] in, represents the penalty coefficient factor; t ij Indicates drone U i Execute the mission Arrive at the mission T j And the time to complete unloading; ET is the maximum allowed time for delivery.

[0067] (3) Establish constraints: including:

[0068] Delivery task constraints. During drone delivery, one drone can deliver multiple delivery tasks, but each delivery task can only be completed by one drone. The formula is as follows:

[0069] ;

[0070] Load constraint. The load capacity of a drone is limited, and the total weight of the mission carried by each drone cannot exceed the drone's load capacity. The formula is as follows:

[0071] ;

[0072] in, q i Indicates i The total load of the drone is less than the maximum load q max ; q j It means the task T j The weight of the cargo required.

[0073] Distance constraint. The flight distance of each logistics drone is limited, and the sum of the flight mission range cannot exceed the maximum flight range of the drone. The formula is as follows:

[0074] .

[0075] (4) Encode and generate the initial chromosome. Figure 1As shown in the figure, the initial encoding method uses integer encoding. Each initial color represents the order of the flight target points of each drone, and the integer of the gene value represents the serial number of each delivery task point. Assume that there are u drones available for the task, and they need to be delivered to v demand points. Label each demand point as 1, 2, ..., v-1, v; if u drones participate in the distribution task, it means that the drone has u times departed from the distribution center, and the distribution center is labeled as v+1; after the demand point label is combined and arranged with u v+1, the basic chromosome is obtained. The genes of the basic chromosome are randomly arranged. In the problem of drone task allocation, the first gene serial number of each chromosome needs to be the starting number of the distribution center. The randomly arranged chromosome genes are tested, and if the genes that do not meet the conditions are detected, the genes need to be repaired. If the starting point position is not the distribution point number, the chromosome is traversed to find the position of the first gene number as the starting gene, and it is exchanged with the first gene. After repair, each chromosome starts from the starting point as the flight route of a drone until the previous position of the next starting gene is cut off.

[0076] (5) Set parameters and use variable neighborhood heuristic optimization to generate the initial population. Set the maximum number of evolution generations M , decision variables, objective function, constraints, population size parameters. Use variable neighborhood heuristic optimization to generate the initial population. After leaving a mission demand point, the drone searches for the next demand point in its circular neighborhood with a variable radius of r. The principle is as follows Figure 2 As shown. Assume that when the drone starts from T1, a circular area is formed with the mission demand point T1 as the center, and the circular area is the neighborhood of T1. The mission demand points contained in the neighborhood are T3, T4 and T7, and the next target point is randomly selected from these three mission demand points. Then repeat the above steps until every mission demand point is traversed. After generating the connection between the basic mission demand points, ensure that the original route sequence sequence remains unchanged, insert the starting node of the number of drones at a random position, and finally generate a chromosome that meets the rules of the encoding method. The neighborhood radius calculation formula is as follows:

[0077] ;

[0078] Among them, k represents the current demand point, d min d avg Respectively represent the minimum distance and average distance between the current point and all other demand points; r(k) max 、r(k) min 、r(k) avg Respectively represent the maximum, minimum and average distances between the current point and other demand points that have not been passed.

[0079] (6) Calculate the selection probability based on individual fitness and select the parent based on the probability. The details are as follows: Calculate the fitness of each chromosome based on the objective function. Sort the chromosomes from small to large in terms of fitness. Calculate the selection probability based on the ranking. P (i), the specific formula is as follows:

[0080] ;

[0081] in, i =1,2… N ; N is the number of individuals; j is the current algebra; M is the maximum evolutionary generation; all individuals get different selection probabilities according to their rankings; and then the parent generation is selected according to the individual's selection probability.

[0082] (7) Perform a mutation crossover operation on the two parents selected in step (6) according to the crossover probability; mutation operation: Figure 3 As shown, during mutation, the positions of the G1 and G2 genes are randomly selected for the two selected parents, and the genes between the selected positions are arranged in reverse order.

[0083] Crossover operation: Figure 4 As shown, the crossover probability is set when crossing p c , after selecting the crossover individual, a number between 0 and 1 is randomly generated. If the random number is greater than or equal to the crossover probability p c , then randomly select two positions G1 and G2 on the gene and invert the gene part between them. If it is less than the crossover probability p c , then find the gene position of G3 adjacent to G1 in another individual, and find two inversions at the original gene to form a new individual; the crossover probability p c The formula is as follows:

[0084] ;

[0085] ;

[0086] in, P c_min is the minimum crossover probability, P c_max is the maximum crossover probability, f is the individual fitness, f max is the maximum value of fitness, f min is the minimum value of fitness, favg is the average fitness of all individuals in the current generation, A , B is a constant.

[0087] (8) The original parent generation and the newly generated next generation are selected as individuals of the next generation according to their fitness ranking.

[0088] (9) Add the offspring obtained in step (8) to the population, and repeat the above operation from step (6) until the number of iterations reaches the maximum. Then the optimal task scheduling solution is obtained.

[0089] In order to evaluate the performance of the present invention, the present invention compares the results with the traditional genetic method. Assume that there are 6 drones with different performances available and 15 mission points to be reached. The evaluation indicators are fitness function value, management cost, transportation cost and delay penalty. The results are shown in Table 1.

[0090] Table 1 Comparison of scheduling results

[0091] ;

[0092] The analysis shows that the optimal, average and maximum fitness values ​​of the present invention are 377.2946, 543.9700 and 876.5671 respectively. The fitness values ​​are 7.13%, 15.94% and 6.49% higher than those of the traditional algorithm respectively. The present invention has improved the fitness function, management cost, transportation cost and delay penalty, which verifies the effectiveness, feasibility and advancement of the present invention.

[0093] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, any one of the methods for scheduling unmanned aerial vehicle flight missions is implemented.

[0094] An embodiment of the present invention further provides a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, any one of the methods for scheduling unmanned aerial vehicle flight missions is implemented.

Claims

1. A method for scheduling UAV flight missions, characterized in that: The following steps are involved: (1) Establish decision variables; the details are as follows: Assume there are n drones, and let the set of drones be U, then , in, ; represents any one of the drones; there are m delivery demand points, each demand point corresponds to a delivery task, and the task set is T, then , in, j =1,2… m represents any one of the delivery tasks; the number of task demand points is greater than the number of drones, that is, m>n; the process of the drone returning to its original position after completing the flight mission is recorded as ; Then the decision variables of the task allocation model are The formula is as follows: ； (2) Set the objective function; the formula is as follows: The objective function F is set to: ； in, , are the weight coefficients of economic cost and delay penalty respectively, and the parameters satisfy ; For economic costs; for delayed punishment; ； ； ； in, Indicates drone The cost of transporting a unit distance at maximum load; Indicates drone Fly from the previous mission point to the mission point The distance length; Indicates drone go through The load-bearing capacity during the period; Indicates drone Maximum load; For drones Own weight; For drones Management cost per unit time; Indicates drone U i Flight speed; ； ； ； in, represents the penalty coefficient factor; t ij Indicates drone U i Execute the mission Arrive at the mission T j And the time to complete unloading; ET is the maximum allowed time for delivery; (3) Establish constraints; the constraints include: Delivery task constraints: ； Load constraints: ； in, q i Indicates i The total load of the drone is less than the maximum load q max ; q j It means the task T j The weight of the cargo required; Distance Constraint: ； (4) Encoding and generating initial chromosomes; (5) Set parameters and use variable neighborhood heuristic optimization to generate the initial population; among them, set the maximum number of evolution generations M , decision variables, objective function, constraints, population size parameters; (6) Calculate the selection probability based on individual fitness and select the parent generation based on the probability; (7) Perform a mutation crossover operation on the two randomly selected parents according to the size of the adaptive crossover probability; (8) The original parent generation and the newly generated next generation are selected as individuals of the next generation according to the fitness ranking; (9) Add the offspring obtained in step (8) to the population and repeat steps (6)-(8) until the number of iterations reaches the maximum.

2. A method for scheduling UAV flight missions according to claim 1, characterized in that: Step (6) is as follows: Calculate the fitness of each chromosome according to the objective function, sort them from small to large in fitness, and calculate the selection probability according to the ranking. P (i), the formula is as follows: ； in, i =1,2… N ; N is the number of individuals; j is the current algebra; M is the maximum evolutionary generation; all individuals get different selection probabilities according to their rankings; and then the parent generation is selected according to the individual's selection probability.

3. The method for scheduling UAV flight missions according to claim 1, characterized in that: Step (7) is as follows: Set the crossover probability p c , after selecting the crossover individual, a number between 0 and 1 is randomly generated. If the random number is greater than or equal to the crossover probability p c , then randomly select two positions G1 and G2 on the gene and invert the gene part between them. If it is less than the crossover probability p c , then find the gene position of G3 adjacent to G1 in another individual, and find the gene between the two inversion points at the original gene to form a new individual; Crossover probability p c The formula is as follows: ； ； in, P c_min is the minimum crossover probability, P c_max is the maximum crossover probability, f is the individual fitness, f max is the maximum value of fitness, f min is the minimum value of fitness, f avg is the average fitness of all individuals in the current generation, A , B is a constant.

4. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into the processor, a method for scheduling UAV flight missions according to any one of claims 1 to 3 is implemented.

5. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a method for scheduling UAV flight missions according to any one of claims 1 to 3 is implemented.

Images